- Email: [email protected]
Understanding the factors affecting the compressive testing of unidirectional carbon fibre composites
Composites: Part B 38 (2007) 481–487 www.elsevier.com/locate/compositesb
Understanding the factors affecting the compressive testing of unidirectional carbon fibre composites Charlene A. Squires b
a,*
, Keith H. Netting a, Alan R. Chambers
b
a Gurit, St Cross Business Park, Newport, Isle of Wight, PO30 5WU, UK School of Engineering Sciences, University of Southampton, Highfield, Southampton, SO17 1BJ, UK
Received 8 May 2006; received in revised form 20 July 2006; accepted 11 August 2006 Available online 1 November 2006
Abstract An investigation was conducted to establish the effects of specimen preparation and configuration on the measured compressive strength of unidirectional carbon fibre. The compressive strength was determined through ASTM D 695 M [ASTM D 695 M, Standard test method for compressive properties of rigid plastics. http://www.astm.org] test method. Specimens conforming to this standard were produced with different thickness, edge and surface preparation. Optical and electron (SEM) microscopical techniques were used to assess initiation of failure and to quantify the damage encountered by the compressive test specimen. The findings correlated well with the mechanical test results, additionally; from the mechanical testing there was significant evidence to suggest that the failure mechanism is dependent upon the quality of the preparation of the test specimen. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: A. Carbon fibre; B. Mechanical properties; C. Mechanical testing
1. Introduction Composite materials offer many exceptional properties that are difficult, or impossible, to match with traditional materials, such as steel, aluminium and wood. Today, composites are used in almost every dynamic, high performance structure whether on land, at sea or in the air. Unfortunately the performance of composites is significantly lower under compression than tensile loading and this is preventing more widespread usage in highly loaded structural applications. The nature of compressive failure in unidirectional composite laminates has been examined for more than three decades. Orringer [2], Shuart [3], Hahn et al. [4] and more recently Anthoine et al. [5], have presented reviews of compressive failure in these laminates. It is accepted that the failure process may involve both elastic and plastic microbuckling matrix failure and fibre fracture. Kink bands *
Corresponding author. Tel.: +44 1983828433. E-mail address: [email protected] (C.A. Squires).
1359-8368/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.compositesb.2006.08.002
formed as a result of in plane buckling may also occur. Compressive failure is matrix dominated and hence improvements in the compressive properties of the resin matrix can be expected to improve the compressive properties of the composite [6]. The performance of unidirectional composites is very dependent on the fibre alignment with respect to the applied load. It has been reported that initial fibre misalignments of the order of 1.5–2°, significantly reduce the compressive strength [7]. Thus accurate alignment of fibres with an absence of waviness is critical to the performance under compression. Compressive data is more difficult to generate than tensile data for carbon fibre reinforced composites. The material has a high ratio of compressive strength (in the direction of the fibres), to shear strength (in the planes parallel to the fibres). This causes problems with shear load input. For pure compression loading, transverse stresses are induced in the ends of the specimen due to Poisson deformation and this can cause brushing at the ends of the specimen. Instability is also a problem and is
482
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
exacerbated by the low out-of-plane shear modulus of the composite. In the axial compression of unidirectional composites three basic failure modes can be observed [8]; local buckling of fibres (where production variations such as fibre waviness or non uniform fibre spacing can influence compressive strength), transverse rupture of the composite (due to differences in Poisson’s ratios of the material constituents and non uniform distribution of transverse strains over the specimen length), and failure in compression (shearing of the fibres at an angle of 45° with no local buckling of the fibres). These principal modes of failure can be accompanied by a series of other phenomena:
Inelastic and non-linear behaviour of fibres and matrix. Interlaminar stresses. Surface ply separation. Overall loss of stability.
Different combinations of all these phenomena can make it very difficult to establish the failure mode or obtain consistent results even with the same material and test procedure. Fig. 1 shows the acceptable failure modes when testing to BS EN ISO 14126: 1999 [9]. The failure mechanisms and measured properties will obviously primarily depend upon the material but will also be influenced by the construction of the test piece. The dimensions of the test piece are defined by the relevant standard and a thickness of 2 mm is recommended. Edge preparation, surface preparation and tabbing are not defined by the standards and no recommendations made. Lee and Soutis [10] have demonstrated that the compressive failure strength decreases with thickness and that the failure mechanism was dependent on thickness. Little research has been conducted on edge and surface preparation. However, Odom and Adams [11] have proposed a sequence of events to describe transverse, split transverse and branched transverse failures. They show that failure initiates at or near a free edge of the specimen and hence imply that edge preparation could be an important factor in determining the incidence of these failures. Failure will occur at the lowest possible stress and in the corresponding failure mode. The range of possible failure strengths implies that the ultimate compressive strength of a composite is not a precise term, but primarily one of
Fig. 1. Acceptable failure mechanisms as stated by BS EN ISO 14126:1999 [8].
definition. Under these circumstances it is hardly surprising that the results from compressive testing are not viewed with great confidence by industry and that scatter within a batch of specimens is high. In interpreting and accepting the results of compression tests the failure mode is an important consideration [12]. The compressive properties of composites are poor in comparison with their tensile properties and ideally should not be subjected to compression. However, in many applications such as wind turbines the loading is complex and elements of compression and flexure are unavoidable. Industry therefore requires reliable data on which to base their designs, select materials and perform structural calculations. There is, therefore, a need to understand the factors, which affect the results and variability of compression testing. The objectives of this report are to optimise the test preparation method and specimen configuration, over and above the method specified in ASTM D 695 M [1], by understanding some of the sources of variation in this test. Three areas of the compressive test specimen were focused upon, these are: thickness of the compressive test specimen, edge quality and also the surface preparation methods for secondary adhesion. 2. Experimental procedure 2.1. Material and specimen fabrication The material that was used in this study was Toray T600-50C unidirectional carbon/epoxy composite (Gurit). The test specimens were manufactured and tested according to ASTM D 695 M [1], Fig. 2 shows the compressive rig used with a specimen in position. Test specimens were manufactured from 2, 3, 4, and 5 plies of T600-50C. The preparation of a test specimen includes the process of cutting the test specimens to size. The method by which this is done is critical to the final quality of the edges of the specimens. There are two cutting methods that have been used, advanced cutting method (method A) and standard cutting method (method B). Both of these techniques use a diamond tipped blade as the cutting medium. They differ in that method A uses a common lubricant mixed with water as the cooling liquid while method B uses water from the mains supply. Method
Fig. 2. ASTM D695 M [1] test fixture with specimen in position.
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
483
Table 1 Summary of compressive and tensile results recorded at Gurit in the last 3 years [12]
Mean strength (MPa) Standard deviation Coefficient of variation (%)
Fig. 3. Example of a compression test specimen.
A is semi-automated and achieves a dimensional accuracy of 0.05 mm. Method B is manual and accuracy and quality of cut are operator dependent. Typically an accuracy of 0.5 mm is achieved. Due to the nature of this compressive test specimen (as shown in Fig. 3) the surface of the carbon fibre component has to be of suitable roughness to guarantee good secondary adhesion for the tabbing. Therefore, the laminate has to undergo a surface preparation technique in order to guarantee the secondary adhesion. The most common surface preparation techniques are the use of peel plies. The three peel plies that are to be used are Release stitch A – coarse mesh, Release stitch G – medium mesh and Release B – fine mesh [13]. Additionally, there are also mechanical abrasion techniques that can be used to prepare the surface for secondary adhesion. In this research, two abrading techniques were used: wet and dry paper and grit blasting. These abrading techniques were applied to the areas on the compressive test panels where it was necessary for the secondary adhesion leaving the gauge length un-abraded. It should be noted that the peel ply covered the whole surface of the compressive test panel and hence affected the surface finish within the gauge length. 2.2. Test procedure The compressive strength testing was completed using a Zwick Z150 static test machine, which has a 250 kN load cell. The compressive strength tests were carried out in accordance with ASTM D695 M [1]. The test machine was calibrated prior to the test programme commencing. The test fixture that was used was the cruciform type as in accordance with the standard. All tests were carried out on samples at 20 °C and at 50 ± 5% room humidity. The specimens were not dried prior to testing.
Compressive data
Tensile data
977.8 118.0 12.1
2119.5 115.9 5.5
The results in Table 1 show, as expected that there is a significant difference between the mean tensile and compressive strengths. With regard to this research the difference in the coefficient of variation (a measure of the scatter) is more relevant. It can be seen that the coefficient of variation from tensile testing is less than half that achieved in compressive testing. If the reasons for the scatter in the compressive tests can be established and the coefficient of variation reduced, industry will have greater confidence in the results and use of compression testing. 3.2. Test facility dependence In order to establish whether, the test facility influenced results, randomly selected test pieces from a laminate but produced by cutting method A were tested in 2 independent test houses. The results are shown in Table 2. It can be seen that the difference in mean compressive strengths was 14% between test facility 1 and test facility 2, and there was a difference of a factor of 2 in the coefficient of variation. As the specimens were randomly selected from the same laminate, it is assumed that the differences in the results are associated with the manner in which the tests were conducted – possibly associated with the alignment of the test pieces. 3.3. The effect of specimen thickness Six panels were manufactured containing 2, 3, 4, and 5 plies of T600-50C unidirectional carbon. These six panels consisted of one panel with two plies, two panels with 3 plies, two panels with 4 plies and one panel with 5 plies, each panel produces a batch of 10 specimens. Fig. 4 contains the results of the 2, 3, 4, and 5 ply. It is evident from the results given in Fig. 4 that the 2-ply laminate was inferior in performance to the 3, 4, and 5 ply specimens due to a low mean failure strength and high variance. Seventy percent of the specimens from the 2-ply batch failed via complex failure mode, which involves buckling of the fibres within the gauge length. The 2 ply laminate was clearly too thin.
3. Results 3.1. Baseline compressive and tensile test data A summary of the compressive and tensile test results that have been compiled at Gurit for T600-50C over a period of 3 years is given in Table 1 [14].
Table 2 Summary of compressive results recorded at the two test houses
Mean strength (MPa) Standard deviation Coefficient of variation (%)
Test facility 1
Test facility 2
898.4 163.9 18.5
1036.7 108.2 9.83
484
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
3.4. The effect of edge quality
Mean Compressive Strength (MPa)
Compressive strength vs Thickness 1600 1400 1200 1000 800 600 400 200 0
2 ply 0
5 ply
1
2
3
Thickness (mm)
Fig. 4. Graph showing the mean compressive strength against 2/3/4/5 ply test specimens.
Considering the 3 and 4 ply laminates, the compressive strengths were similar (Fig. 4). However, the coefficient of variation of the 3 ply laminate was slightly higher (7.14%) than that of the 4 ply (5.51%). The Euler buckling load was calculated for the unidirectional carbon specimens (3 and 4 ply). The values calculated do not take into consideration the effects of the glass triax tabbing (3 ply – 1.6 MPa and 4 ply – 3.5 MPa). These results suggest that the 3-ply was close to its operational limit because of its buckling load, whereas the 4-ply laminate has not reached its buckling limit. As buckling is not an acceptable failure mode in compression testing, there is a greater risk of buckling influencing the results from 3 ply than there is with 4 ply. Buckling is a problem with this type of compression testing, and if the specimen fails by this mode the result will be ignored from the batch of specimens. Increasing the thickness to 5 ply, yielded no further improvements over the 4 ply in either the compressive strength or reducing the coefficient of variation. Additionally, the batch tested that were manufactured from 5 plies, 80% of the specimens failed by end crushing which is an unacceptable failure mode in BS EN ISO 14126:1999 [9] and hence in this analysis the results from these specimens were disregarded. The ASTM D 695 M [1] does not characterise suitable failure modes. In conclusion, the optimum ply thickness for this quality laminate was 4 plies (2 mm). This is consistent with that recommended in standard BS EN ISO 14126:1999 [9] and agrees with the work of Lee and Soutis [10].
Twelve batches of compressive test specimens were manufactured using 4 plies of T600-50C unidirectional carbon. The specimens were aligned prior to cutting. Six of the batches were prepared using advanced cutting method A and six prepared using standard cutting method B. These batches were evaluated on a Talyscan 150 profilometer to quantify the edge roughness that each preparation method produces. Fig. 5 shows the effect of edge roughness on the compressive strength. Method A consistently produced a surface finish of between 4 and 5 lm average surface roughness, whereas cutting method B produced a surface finish in the range 3.5–22 lm. The coefficient of variation for the cutting method A (6.0%) was significantly lower than cutting method B (17.3%). The results show a trend of increasing edge roughness reducing the compressive strength. The compressive strength at 22 lm Ra was only 60% of the 4 lm roughness samples which strongly suggests that edge quality plays an important role in the initiation of compressive failure and hence the results of compressive testing. These test specimens were examined using an optical microscope prior to mechanical testing to visually assess edge condition resulting from the preparation methods. It was found that method A produced a consistent finish free from saw marks and abrasions (Fig. 6). This contrasts with method B which caused saw abrasions and penetration marks as shown in Fig. 7(a) and (b) and in greater detail in the SEM image (Fig. 7(c)). The type of damage caused
Fig. 6. Example of the Method A edge finish.
Mean Compressive Stength (MPa)
Compressive Strength vs Edge Roughness 1400 1200 1000 800 600 400 200 0
0
5
10 15 Mean Edge Roughness (Ra(µ))
20
25
Fig. 5. Graph showing the mean compressive strength against measured edge roughness.
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
485
Fig. 7. (a) and (b) Example of the Method B finish, (c) SEM image of the abrasion in image (b).
was used. Fig. 8 shows the mean compressive strength achieved by the batches against the mean surface roughness recorded using the profilometer. Wet and dry abrasion and grit blasting preparation techniques achieved the smoothest surfaces and the lowest compressive strengths. The failures were clearly tabbing (adhesion) failures caused by the surface roughness being insufficient for secondary adhesion. The compressive tests were essentially testing the construction of the test specimen rather than the compressive strength of the composite. The three different types of peel ply range from coarse (Release stitch A) to fine (Release B) mesh grades, with Release stitch G between A and B. Release B peel ply resulted in at least 50% of the test specimens from each of the two batches failing via adhesion failure. An average surface roughness value of 9 lm is too smooth for consistent secondary adhesion. Release stitch A and Release stitch G provided very similar results, in terms of mean compressive strength and coefficient of variation. There were no adhesion failures with either of these preparation methods. Release stitch A produced an average surface roughness of 14.9 lm
by Method B could have been due either to the alignment of the blade and/or the operator’s control. Overall the coefficient of variation from the mechanical test results correlated well with the Talyscan results (Fig. 5). Method A gave the highest compressive strength results and the lowest variance. The specimen that had the highest edge roughness (prepared using method B) also achieved the lowest compressive strength value. This investigation has shown that edge condition plays a role in the initiation of failure in a compression test. To maximise the failure strength and minimise the coefficient of variation a smooth, defect free edge surface is required. 3.5. The effect of surface preparation
Mean Compressive Strength (MPa)
The five surface preparation techniques that were applied to the T600-50C unidirectional carbon fibre in order to provide a surface suitable for secondary adhesion for the end tabs were: wet and dry sandpaper (Grade 400), grit blasting (aluminium oxide 6040), Release stitch A peel ply, Release B peel ply and Release stitch G peel ply [13]. For the adhesion of the end tabs an epoxy based adhesive 1600 1400 1200 1000 800 600 400
0
2
4
6
8
10
12
14
16
18
Surface Roughness Wet and Dry
Grit Blasting
Release A
Release B
Release G
Fig. 8. Graph showing the mean compressive strength against measured surface roughness.
486
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487 Table 3 Summary of the failure modes that occurred in each batch tested Failure modes
Fig. 9. SEM image of the residue left by Release peel ply A.
compared with Release stitch G that achieved 11.5 lm average surface roughness. The test specimens were analysed using the SEM. The Release stitch A peel ply resulted in a very coarse surface with a texture composed of peaks and troughs. Fig. 9 is an image taken on the SEM of a peak left by the Release stitch A peel ply after removal from the laminate. The deposits were measured to be 0.014 mm. This is a significant thickness considering the specimen is only 2 mm thick and will adversely affect the calculated compressive strength. Release G also produced a textured surface comprising peaks and troughs but was slightly smoother with an average thickness of 0.01 mm.
3.6. Failure modes The failure modes observed during this research are summarised in Table 3. From the edge quality research completed there were very strong indications that the inplane shear failure mechanism (Fig. 10(a)) was a result of poor edge quality. Seventy percent of the specimens failed by this mechanism compared with only 20% of the batch prepared by method A. For cutting method A, the dominant failure mode was through thickness shear failure which is surface initiated. The incidence of through thickness shear failures was greater with release stitch A than with release stitch B. This can be attributed to the greater surface roughness providing potential initiation sites at any point on the surface within the gauge length, resulting from the resin deposits left on the surface. It was interesting to note that with release stitch G, the majority of the through thickness shear failures initiated at the stress concentration at the end of
2 Ply 3 Ply 4 Ply 5 Ply Advanced cutting method Standard cutting method Release stitch A Release B Release stitch G Wet and dry sandpaper Grit blasting
In plane shear (%)
Through thickness shear (%)
Complex (%)
30 20 5 9 20
20 30 70
50 50 25
70
10
70
30
10
70
20
40 40
60
End failure (%)
Adhesion failure (%)
91
60
40
60
25
75
the tab area (Fig. 10(b)). This suggests that with release stitch A, the stress concentrations within the gauge length due to the surface roughness are greater than those at the tab. The final failure mode that indicated a trend was the complex mode of failure or buckling (Fig. 10(c)). There was evidence from the 2-ply laminate that failed by this mode. With the thicker 4 ply samples this mode of failure was observed with the samples, which gave the highest compressive strength results and hence is the preferred failure mode. 4. Discussion The research has demonstrated that the results from compression testing, which include the compressive strength and variance, are strongly influenced by factors concerning the preparation of the samples that are not covered or defined by the relevant standards. It has been shown that the 2 mm thickness of the samples as recommended by the standard and by Lee and Soutis [10] is appropriate. Thinner samples are likely to fail prematurely due to buckling whereas there is no gain either
Fig. 10. (a) In plane shear failure; (b) through thickness shear failure and (c) complex failure.
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
in compressive strength or coefficient of variation from using thicker samples. The quality of the sample cut edges proved to be an important variable because with edges containing defects such as saw abrasions failed at lower loads and batches of samples produced by operator dependent methods showed significantly higher variance in the results. Poor edge preparation induced in plane shear failures. The results indicate that low coefficient of variation can be achieved with an edge roughness of <5 lm. The quality of the free surface is equally important. A surface, that is, too smooth will result in a low compressive strength and high coefficient of variation because of tabbing adhesion failure. Conversely a surface, which is too rough, may result in increased coefficient of variation because of through thickness shear failures initiating from high roughness sites on the free surface. The results suggest that a surface finish in the range 10–13 lm should be sought. The above features influence the mode of failure. With a well prepared specimen, the dominant failure mechanism is complex (Fig. 1) which involves the load being carried by the fibres and the matrix keeping the fibres vertical in plane. Subject to thickness, if this failure mechanism is observed then the result can be treated with confidence. 5. Conclusions From the results obtained through this research, it is clear that the results of compression tests are influenced by: (1) (2) (3) (4)
Edge preparation. Surface preparation. Test piece thickness. Test piece alignment.
487
To achieve reliable compressive strengths from the standard compressive tests, which reflect the quality of the composite, a consistent smooth specimen edge is required and a free surface of roughness sufficient to promote tabbing adhesion without causing premature failure from roughness features on the free surface within the gauge length. References [1] ASTM D 695 M, Standard test method for compressive properties of rigid plastics. http://www.astm.org. [2] Orringer, Compressive behaviour of fibre composites. AFOSR Technical Report, TR-71-3098, 1971. [3] Shuart MJ. Short wavelength buckling and shear failures for compression loaded laminates. NASA TM 87640, 1985. [4] Hahn HT, Sohi M, Moon S. Compression failure mechanisms of composite structures. Washington University St Louis Centre for Composites and Research; 1986. [5] Anthoine O, Grandidier JC, Daridon L. Puer compression testing of advanced fibre composites. Compos Sci Technol 1998;58:735–40. [6] Hodgkinson JM. Mechanical testing of advanced fibre composites. 2000; ISBN 1 85573 312 9. [7] Hsiao HM, Daniel IM. Effect of fibre waviness on stiffness and strength reduction of unidirectional composites under compressive loading. Compos Sci Technol 1996;56:581–93. [8] Swift DG. Elastic moduli of fibrous composites containing misaligned fibres. J Phys D: Appl Phys 1975;8:223. [9] BS EN ISO 14126:1999. [10] Lee J, Soutis C. Thickness effect on the compressive strength of T600/ 924C carbon-fibre epoxy laminates. Compos Part A 2005;36:213–27. [11] Odom EM, Adams DF. Failure modes of unidirectional carbon/ epoxy composite compression specimens. Composites 1990;21: 289–96. [12] Whitney JM. Reflections on the development of test methods for advanced composite materials. In: Proceedings of 10th conference ‘composite materials: testing and design’, 1992, p. 7–16. [13] Tygravac http://www.airtechonline.com. [14] Gurit Composites Mechanical Testing Database.
Understanding the factors affecting the compressive testing of unidirectional carbon fibre composites Charlene A. Squires b
a,*
, Keith H. Netting a, Alan R. Chambers
b
a Gurit, St Cross Business Park, Newport, Isle of Wight, PO30 5WU, UK School of Engineering Sciences, University of Southampton, Highfield, Southampton, SO17 1BJ, UK
Received 8 May 2006; received in revised form 20 July 2006; accepted 11 August 2006 Available online 1 November 2006
Abstract An investigation was conducted to establish the effects of specimen preparation and configuration on the measured compressive strength of unidirectional carbon fibre. The compressive strength was determined through ASTM D 695 M [ASTM D 695 M, Standard test method for compressive properties of rigid plastics. http://www.astm.org] test method. Specimens conforming to this standard were produced with different thickness, edge and surface preparation. Optical and electron (SEM) microscopical techniques were used to assess initiation of failure and to quantify the damage encountered by the compressive test specimen. The findings correlated well with the mechanical test results, additionally; from the mechanical testing there was significant evidence to suggest that the failure mechanism is dependent upon the quality of the preparation of the test specimen. Ó 2006 Elsevier Ltd. All rights reserved. Keywords: A. Carbon fibre; B. Mechanical properties; C. Mechanical testing
1. Introduction Composite materials offer many exceptional properties that are difficult, or impossible, to match with traditional materials, such as steel, aluminium and wood. Today, composites are used in almost every dynamic, high performance structure whether on land, at sea or in the air. Unfortunately the performance of composites is significantly lower under compression than tensile loading and this is preventing more widespread usage in highly loaded structural applications. The nature of compressive failure in unidirectional composite laminates has been examined for more than three decades. Orringer [2], Shuart [3], Hahn et al. [4] and more recently Anthoine et al. [5], have presented reviews of compressive failure in these laminates. It is accepted that the failure process may involve both elastic and plastic microbuckling matrix failure and fibre fracture. Kink bands *
Corresponding author. Tel.: +44 1983828433. E-mail address: [email protected] (C.A. Squires).
1359-8368/$ - see front matter Ó 2006 Elsevier Ltd. All rights reserved. doi:10.1016/j.compositesb.2006.08.002
formed as a result of in plane buckling may also occur. Compressive failure is matrix dominated and hence improvements in the compressive properties of the resin matrix can be expected to improve the compressive properties of the composite [6]. The performance of unidirectional composites is very dependent on the fibre alignment with respect to the applied load. It has been reported that initial fibre misalignments of the order of 1.5–2°, significantly reduce the compressive strength [7]. Thus accurate alignment of fibres with an absence of waviness is critical to the performance under compression. Compressive data is more difficult to generate than tensile data for carbon fibre reinforced composites. The material has a high ratio of compressive strength (in the direction of the fibres), to shear strength (in the planes parallel to the fibres). This causes problems with shear load input. For pure compression loading, transverse stresses are induced in the ends of the specimen due to Poisson deformation and this can cause brushing at the ends of the specimen. Instability is also a problem and is
482
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
exacerbated by the low out-of-plane shear modulus of the composite. In the axial compression of unidirectional composites three basic failure modes can be observed [8]; local buckling of fibres (where production variations such as fibre waviness or non uniform fibre spacing can influence compressive strength), transverse rupture of the composite (due to differences in Poisson’s ratios of the material constituents and non uniform distribution of transverse strains over the specimen length), and failure in compression (shearing of the fibres at an angle of 45° with no local buckling of the fibres). These principal modes of failure can be accompanied by a series of other phenomena:
Inelastic and non-linear behaviour of fibres and matrix. Interlaminar stresses. Surface ply separation. Overall loss of stability.
Different combinations of all these phenomena can make it very difficult to establish the failure mode or obtain consistent results even with the same material and test procedure. Fig. 1 shows the acceptable failure modes when testing to BS EN ISO 14126: 1999 [9]. The failure mechanisms and measured properties will obviously primarily depend upon the material but will also be influenced by the construction of the test piece. The dimensions of the test piece are defined by the relevant standard and a thickness of 2 mm is recommended. Edge preparation, surface preparation and tabbing are not defined by the standards and no recommendations made. Lee and Soutis [10] have demonstrated that the compressive failure strength decreases with thickness and that the failure mechanism was dependent on thickness. Little research has been conducted on edge and surface preparation. However, Odom and Adams [11] have proposed a sequence of events to describe transverse, split transverse and branched transverse failures. They show that failure initiates at or near a free edge of the specimen and hence imply that edge preparation could be an important factor in determining the incidence of these failures. Failure will occur at the lowest possible stress and in the corresponding failure mode. The range of possible failure strengths implies that the ultimate compressive strength of a composite is not a precise term, but primarily one of
Fig. 1. Acceptable failure mechanisms as stated by BS EN ISO 14126:1999 [8].
definition. Under these circumstances it is hardly surprising that the results from compressive testing are not viewed with great confidence by industry and that scatter within a batch of specimens is high. In interpreting and accepting the results of compression tests the failure mode is an important consideration [12]. The compressive properties of composites are poor in comparison with their tensile properties and ideally should not be subjected to compression. However, in many applications such as wind turbines the loading is complex and elements of compression and flexure are unavoidable. Industry therefore requires reliable data on which to base their designs, select materials and perform structural calculations. There is, therefore, a need to understand the factors, which affect the results and variability of compression testing. The objectives of this report are to optimise the test preparation method and specimen configuration, over and above the method specified in ASTM D 695 M [1], by understanding some of the sources of variation in this test. Three areas of the compressive test specimen were focused upon, these are: thickness of the compressive test specimen, edge quality and also the surface preparation methods for secondary adhesion. 2. Experimental procedure 2.1. Material and specimen fabrication The material that was used in this study was Toray T600-50C unidirectional carbon/epoxy composite (Gurit). The test specimens were manufactured and tested according to ASTM D 695 M [1], Fig. 2 shows the compressive rig used with a specimen in position. Test specimens were manufactured from 2, 3, 4, and 5 plies of T600-50C. The preparation of a test specimen includes the process of cutting the test specimens to size. The method by which this is done is critical to the final quality of the edges of the specimens. There are two cutting methods that have been used, advanced cutting method (method A) and standard cutting method (method B). Both of these techniques use a diamond tipped blade as the cutting medium. They differ in that method A uses a common lubricant mixed with water as the cooling liquid while method B uses water from the mains supply. Method
Fig. 2. ASTM D695 M [1] test fixture with specimen in position.
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
483
Table 1 Summary of compressive and tensile results recorded at Gurit in the last 3 years [12]
Mean strength (MPa) Standard deviation Coefficient of variation (%)
Fig. 3. Example of a compression test specimen.
A is semi-automated and achieves a dimensional accuracy of 0.05 mm. Method B is manual and accuracy and quality of cut are operator dependent. Typically an accuracy of 0.5 mm is achieved. Due to the nature of this compressive test specimen (as shown in Fig. 3) the surface of the carbon fibre component has to be of suitable roughness to guarantee good secondary adhesion for the tabbing. Therefore, the laminate has to undergo a surface preparation technique in order to guarantee the secondary adhesion. The most common surface preparation techniques are the use of peel plies. The three peel plies that are to be used are Release stitch A – coarse mesh, Release stitch G – medium mesh and Release B – fine mesh [13]. Additionally, there are also mechanical abrasion techniques that can be used to prepare the surface for secondary adhesion. In this research, two abrading techniques were used: wet and dry paper and grit blasting. These abrading techniques were applied to the areas on the compressive test panels where it was necessary for the secondary adhesion leaving the gauge length un-abraded. It should be noted that the peel ply covered the whole surface of the compressive test panel and hence affected the surface finish within the gauge length. 2.2. Test procedure The compressive strength testing was completed using a Zwick Z150 static test machine, which has a 250 kN load cell. The compressive strength tests were carried out in accordance with ASTM D695 M [1]. The test machine was calibrated prior to the test programme commencing. The test fixture that was used was the cruciform type as in accordance with the standard. All tests were carried out on samples at 20 °C and at 50 ± 5% room humidity. The specimens were not dried prior to testing.
Compressive data
Tensile data
977.8 118.0 12.1
2119.5 115.9 5.5
The results in Table 1 show, as expected that there is a significant difference between the mean tensile and compressive strengths. With regard to this research the difference in the coefficient of variation (a measure of the scatter) is more relevant. It can be seen that the coefficient of variation from tensile testing is less than half that achieved in compressive testing. If the reasons for the scatter in the compressive tests can be established and the coefficient of variation reduced, industry will have greater confidence in the results and use of compression testing. 3.2. Test facility dependence In order to establish whether, the test facility influenced results, randomly selected test pieces from a laminate but produced by cutting method A were tested in 2 independent test houses. The results are shown in Table 2. It can be seen that the difference in mean compressive strengths was 14% between test facility 1 and test facility 2, and there was a difference of a factor of 2 in the coefficient of variation. As the specimens were randomly selected from the same laminate, it is assumed that the differences in the results are associated with the manner in which the tests were conducted – possibly associated with the alignment of the test pieces. 3.3. The effect of specimen thickness Six panels were manufactured containing 2, 3, 4, and 5 plies of T600-50C unidirectional carbon. These six panels consisted of one panel with two plies, two panels with 3 plies, two panels with 4 plies and one panel with 5 plies, each panel produces a batch of 10 specimens. Fig. 4 contains the results of the 2, 3, 4, and 5 ply. It is evident from the results given in Fig. 4 that the 2-ply laminate was inferior in performance to the 3, 4, and 5 ply specimens due to a low mean failure strength and high variance. Seventy percent of the specimens from the 2-ply batch failed via complex failure mode, which involves buckling of the fibres within the gauge length. The 2 ply laminate was clearly too thin.
3. Results 3.1. Baseline compressive and tensile test data A summary of the compressive and tensile test results that have been compiled at Gurit for T600-50C over a period of 3 years is given in Table 1 [14].
Table 2 Summary of compressive results recorded at the two test houses
Mean strength (MPa) Standard deviation Coefficient of variation (%)
Test facility 1
Test facility 2
898.4 163.9 18.5
1036.7 108.2 9.83
484
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
3.4. The effect of edge quality
Mean Compressive Strength (MPa)
Compressive strength vs Thickness 1600 1400 1200 1000 800 600 400 200 0
2 ply 0
5 ply
1
2
3
Thickness (mm)
Fig. 4. Graph showing the mean compressive strength against 2/3/4/5 ply test specimens.
Considering the 3 and 4 ply laminates, the compressive strengths were similar (Fig. 4). However, the coefficient of variation of the 3 ply laminate was slightly higher (7.14%) than that of the 4 ply (5.51%). The Euler buckling load was calculated for the unidirectional carbon specimens (3 and 4 ply). The values calculated do not take into consideration the effects of the glass triax tabbing (3 ply – 1.6 MPa and 4 ply – 3.5 MPa). These results suggest that the 3-ply was close to its operational limit because of its buckling load, whereas the 4-ply laminate has not reached its buckling limit. As buckling is not an acceptable failure mode in compression testing, there is a greater risk of buckling influencing the results from 3 ply than there is with 4 ply. Buckling is a problem with this type of compression testing, and if the specimen fails by this mode the result will be ignored from the batch of specimens. Increasing the thickness to 5 ply, yielded no further improvements over the 4 ply in either the compressive strength or reducing the coefficient of variation. Additionally, the batch tested that were manufactured from 5 plies, 80% of the specimens failed by end crushing which is an unacceptable failure mode in BS EN ISO 14126:1999 [9] and hence in this analysis the results from these specimens were disregarded. The ASTM D 695 M [1] does not characterise suitable failure modes. In conclusion, the optimum ply thickness for this quality laminate was 4 plies (2 mm). This is consistent with that recommended in standard BS EN ISO 14126:1999 [9] and agrees with the work of Lee and Soutis [10].
Twelve batches of compressive test specimens were manufactured using 4 plies of T600-50C unidirectional carbon. The specimens were aligned prior to cutting. Six of the batches were prepared using advanced cutting method A and six prepared using standard cutting method B. These batches were evaluated on a Talyscan 150 profilometer to quantify the edge roughness that each preparation method produces. Fig. 5 shows the effect of edge roughness on the compressive strength. Method A consistently produced a surface finish of between 4 and 5 lm average surface roughness, whereas cutting method B produced a surface finish in the range 3.5–22 lm. The coefficient of variation for the cutting method A (6.0%) was significantly lower than cutting method B (17.3%). The results show a trend of increasing edge roughness reducing the compressive strength. The compressive strength at 22 lm Ra was only 60% of the 4 lm roughness samples which strongly suggests that edge quality plays an important role in the initiation of compressive failure and hence the results of compressive testing. These test specimens were examined using an optical microscope prior to mechanical testing to visually assess edge condition resulting from the preparation methods. It was found that method A produced a consistent finish free from saw marks and abrasions (Fig. 6). This contrasts with method B which caused saw abrasions and penetration marks as shown in Fig. 7(a) and (b) and in greater detail in the SEM image (Fig. 7(c)). The type of damage caused
Fig. 6. Example of the Method A edge finish.
Mean Compressive Stength (MPa)
Compressive Strength vs Edge Roughness 1400 1200 1000 800 600 400 200 0
0
5
10 15 Mean Edge Roughness (Ra(µ))
20
25
Fig. 5. Graph showing the mean compressive strength against measured edge roughness.
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
485
Fig. 7. (a) and (b) Example of the Method B finish, (c) SEM image of the abrasion in image (b).
was used. Fig. 8 shows the mean compressive strength achieved by the batches against the mean surface roughness recorded using the profilometer. Wet and dry abrasion and grit blasting preparation techniques achieved the smoothest surfaces and the lowest compressive strengths. The failures were clearly tabbing (adhesion) failures caused by the surface roughness being insufficient for secondary adhesion. The compressive tests were essentially testing the construction of the test specimen rather than the compressive strength of the composite. The three different types of peel ply range from coarse (Release stitch A) to fine (Release B) mesh grades, with Release stitch G between A and B. Release B peel ply resulted in at least 50% of the test specimens from each of the two batches failing via adhesion failure. An average surface roughness value of 9 lm is too smooth for consistent secondary adhesion. Release stitch A and Release stitch G provided very similar results, in terms of mean compressive strength and coefficient of variation. There were no adhesion failures with either of these preparation methods. Release stitch A produced an average surface roughness of 14.9 lm
by Method B could have been due either to the alignment of the blade and/or the operator’s control. Overall the coefficient of variation from the mechanical test results correlated well with the Talyscan results (Fig. 5). Method A gave the highest compressive strength results and the lowest variance. The specimen that had the highest edge roughness (prepared using method B) also achieved the lowest compressive strength value. This investigation has shown that edge condition plays a role in the initiation of failure in a compression test. To maximise the failure strength and minimise the coefficient of variation a smooth, defect free edge surface is required. 3.5. The effect of surface preparation
Mean Compressive Strength (MPa)
The five surface preparation techniques that were applied to the T600-50C unidirectional carbon fibre in order to provide a surface suitable for secondary adhesion for the end tabs were: wet and dry sandpaper (Grade 400), grit blasting (aluminium oxide 6040), Release stitch A peel ply, Release B peel ply and Release stitch G peel ply [13]. For the adhesion of the end tabs an epoxy based adhesive 1600 1400 1200 1000 800 600 400
0
2
4
6
8
10
12
14
16
18
Surface Roughness Wet and Dry
Grit Blasting
Release A
Release B
Release G
Fig. 8. Graph showing the mean compressive strength against measured surface roughness.
486
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487 Table 3 Summary of the failure modes that occurred in each batch tested Failure modes
Fig. 9. SEM image of the residue left by Release peel ply A.
compared with Release stitch G that achieved 11.5 lm average surface roughness. The test specimens were analysed using the SEM. The Release stitch A peel ply resulted in a very coarse surface with a texture composed of peaks and troughs. Fig. 9 is an image taken on the SEM of a peak left by the Release stitch A peel ply after removal from the laminate. The deposits were measured to be 0.014 mm. This is a significant thickness considering the specimen is only 2 mm thick and will adversely affect the calculated compressive strength. Release G also produced a textured surface comprising peaks and troughs but was slightly smoother with an average thickness of 0.01 mm.
3.6. Failure modes The failure modes observed during this research are summarised in Table 3. From the edge quality research completed there were very strong indications that the inplane shear failure mechanism (Fig. 10(a)) was a result of poor edge quality. Seventy percent of the specimens failed by this mechanism compared with only 20% of the batch prepared by method A. For cutting method A, the dominant failure mode was through thickness shear failure which is surface initiated. The incidence of through thickness shear failures was greater with release stitch A than with release stitch B. This can be attributed to the greater surface roughness providing potential initiation sites at any point on the surface within the gauge length, resulting from the resin deposits left on the surface. It was interesting to note that with release stitch G, the majority of the through thickness shear failures initiated at the stress concentration at the end of
2 Ply 3 Ply 4 Ply 5 Ply Advanced cutting method Standard cutting method Release stitch A Release B Release stitch G Wet and dry sandpaper Grit blasting
In plane shear (%)
Through thickness shear (%)
Complex (%)
30 20 5 9 20
20 30 70
50 50 25
70
10
70
30
10
70
20
40 40
60
End failure (%)
Adhesion failure (%)
91
60
40
60
25
75
the tab area (Fig. 10(b)). This suggests that with release stitch A, the stress concentrations within the gauge length due to the surface roughness are greater than those at the tab. The final failure mode that indicated a trend was the complex mode of failure or buckling (Fig. 10(c)). There was evidence from the 2-ply laminate that failed by this mode. With the thicker 4 ply samples this mode of failure was observed with the samples, which gave the highest compressive strength results and hence is the preferred failure mode. 4. Discussion The research has demonstrated that the results from compression testing, which include the compressive strength and variance, are strongly influenced by factors concerning the preparation of the samples that are not covered or defined by the relevant standards. It has been shown that the 2 mm thickness of the samples as recommended by the standard and by Lee and Soutis [10] is appropriate. Thinner samples are likely to fail prematurely due to buckling whereas there is no gain either
Fig. 10. (a) In plane shear failure; (b) through thickness shear failure and (c) complex failure.
C.A. Squires et al. / Composites: Part B 38 (2007) 481–487
in compressive strength or coefficient of variation from using thicker samples. The quality of the sample cut edges proved to be an important variable because with edges containing defects such as saw abrasions failed at lower loads and batches of samples produced by operator dependent methods showed significantly higher variance in the results. Poor edge preparation induced in plane shear failures. The results indicate that low coefficient of variation can be achieved with an edge roughness of <5 lm. The quality of the free surface is equally important. A surface, that is, too smooth will result in a low compressive strength and high coefficient of variation because of tabbing adhesion failure. Conversely a surface, which is too rough, may result in increased coefficient of variation because of through thickness shear failures initiating from high roughness sites on the free surface. The results suggest that a surface finish in the range 10–13 lm should be sought. The above features influence the mode of failure. With a well prepared specimen, the dominant failure mechanism is complex (Fig. 1) which involves the load being carried by the fibres and the matrix keeping the fibres vertical in plane. Subject to thickness, if this failure mechanism is observed then the result can be treated with confidence. 5. Conclusions From the results obtained through this research, it is clear that the results of compression tests are influenced by: (1) (2) (3) (4)
Edge preparation. Surface preparation. Test piece thickness. Test piece alignment.
487
To achieve reliable compressive strengths from the standard compressive tests, which reflect the quality of the composite, a consistent smooth specimen edge is required and a free surface of roughness sufficient to promote tabbing adhesion without causing premature failure from roughness features on the free surface within the gauge length. References [1] ASTM D 695 M, Standard test method for compressive properties of rigid plastics. http://www.astm.org. [2] Orringer, Compressive behaviour of fibre composites. AFOSR Technical Report, TR-71-3098, 1971. [3] Shuart MJ. Short wavelength buckling and shear failures for compression loaded laminates. NASA TM 87640, 1985. [4] Hahn HT, Sohi M, Moon S. Compression failure mechanisms of composite structures. Washington University St Louis Centre for Composites and Research; 1986. [5] Anthoine O, Grandidier JC, Daridon L. Puer compression testing of advanced fibre composites. Compos Sci Technol 1998;58:735–40. [6] Hodgkinson JM. Mechanical testing of advanced fibre composites. 2000; ISBN 1 85573 312 9. [7] Hsiao HM, Daniel IM. Effect of fibre waviness on stiffness and strength reduction of unidirectional composites under compressive loading. Compos Sci Technol 1996;56:581–93. [8] Swift DG. Elastic moduli of fibrous composites containing misaligned fibres. J Phys D: Appl Phys 1975;8:223. [9] BS EN ISO 14126:1999. [10] Lee J, Soutis C. Thickness effect on the compressive strength of T600/ 924C carbon-fibre epoxy laminates. Compos Part A 2005;36:213–27. [11] Odom EM, Adams DF. Failure modes of unidirectional carbon/ epoxy composite compression specimens. Composites 1990;21: 289–96. [12] Whitney JM. Reflections on the development of test methods for advanced composite materials. In: Proceedings of 10th conference ‘composite materials: testing and design’, 1992, p. 7–16. [13] Tygravac http://www.airtechonline.com. [14] Gurit Composites Mechanical Testing Database.